Epidemiology: Basics of Study Design, Measures, and Limitations LSU Neal Simonsen Health Sciences Center Defining Epidemiology the study of the distribution and determinants of health-related states or events in specified populations and the application of this study to control of health problems. --Last, 1988, The Dictionary of Epidemiology School of Public Health & Stanley S. Scott Cancer Center Another Definition of Epidemiology The science of making the obvious obscure. --Anonymous Epidemiologist Epidemiology Defining Itself Etymological derivation: : From the Greek Epi on/upon + demos the people + logos theory or study of Characteristics of Epidemiology Concerned with the frequencies and types of illnesses and injuries in groups of people and the factors that influence their distribution. 1
Characteristics (Continued) This implies that disease is NOT randomly distributed throughout a population, but rather that subgroups differ in the frequency of different diseases. Knowledge of this uneven distribution can be used to investigate causal factors and thus to lay the groundwork for programs of prevention and control. Can similarly be used to study consequences of different treatments Prevalence vs. Incidence rates Prevalence Proportion of persons in a population who have a particular status (presence of disease or some other health-related condition) at either 1) A specified point in time 2) A specified period in time Point vs. period prevalence Incidence Incidence The rate of new occurrences of a condition in a population = (New occurrences of a condition [=event] during a specified time period / the population during that same specified time period) Prevalence vs. Incidence rates Prevalence (continued) Amount of disease prevailing in a population at a given time or within a given period What if we are interested in how quickly new cases are developing in a population? What if the condition of interest lasts a long time (years)? E.g., osteoarthritis Study Designs In Epidemiology Basic Considerations; Fundamental Designs Exposure OR Genetic Background OR Combination of Both? Causation? or Other Outcome Association How do we know if an observed association reflects a causal relationship? 2
Exploring Etiology Dose/Treatment Experimental Study Design Association Cancer Exposure assigned Not exposed nonoccurrence OK Environmental Exposure? Cancer Unethical to perform experiments on people if exposure is harmful Not OK http://www.epiet.org/course/presentations/presentations%202003/13-%20cohort&case%20control/13-%20cohort&case%20control_files/frame.htm The next step in determining causation: Conducting Studies in Human Populations Observational Epidemiology often key here. Allows capitalization on natural or unplanned experiments. Take advantage of groups who have been exposed for non-study purposes. Clinical observations (case series) Ecological or Cross-Sectional studies Case - control studies Cohort studies Randomized trials* *(if potential beneficial intervention identified) Ecologic Study Units of analysis are populations or groups of people,, rather than individuals. Often exploit pre-existing existing data collected for other purposes Efficient and economical design Incidence Ratio per 100,000 Women Correlation between dietary fat intake and breast cancer by country. 250 200 150 100 Romania Hong Kong 50 Japan USA Switzerland Canada Fed. Repub. Italy Of Germany Israel UK Denmark Sweden France New Zealand Australia Norway Finland Yugoslavia Spain Poland Hungary 0 0 600 800 1000 1200 1400 1600 Prentice RL, Kakar F, Hursting S, et al: Aspects of Per Capita Supply of Fat Calories the rationale for the Women s Health Trial. J Natl Cancer Inst 80:802-814, 1988.) 3
Key potential limitation: The ecologic fallacy Attributing to members of a group characteristics that they do not possess as individuals E.g., only know average values of fat consumption by country Don t t know if individuals with breast cancer had higher fat intake Cross-sectional Study Draw sample from population of interest at particular time Identify cases and non-cases of disease Measure characteristics (exposures) Examine associations between characteristics and disease Example: Is stress associated with symptoms of TMD? Random sample of population (N=680) Interviewed re: symptoms of TMD (pain, joint sounds, limited opening) Measure of life stress Stress and TMD Percent reporting frequent stress: Those with TMD symptoms = 56% Those without symptoms = 21% P<0.05 Cross-sectional studies: Cohort Study Can assess associations Cannot establish correct temporal relationship for inferring causation Why? Factor and disease measured at same point in time Not 4
The Cohort Concept Following the cohort through time t 0 t 1 t 2 to the end of the study period. First select Analytical Design of a Cohort Study Then follow to see whether does Totals develops not develop Incidence Rates of a a b a + b a + b c Not exposed c d c + d c + d End of Follow-Up Relative Risk Risk in exposed Risk in non-exposed Relative Risk Calculation for Cohort Study group group Incidence CHD per 1,000 CHD Totals per Year New New incidents incidents Total Total population population Smoke 84 2,916 3,000 28.0 (= 84/3000) Cigarettes ( x 1,000 ) Do not smoke 87 4,913 5,000 17.4 (= 87/5000) Cigarettes ( x 1,000 ) RR = 28.0 / 17.4 = 1.61 5
Interpreting RR of a If RR = 1 Risk in exposed equal to risk in unexposed (no association) If RR > 1 Risk in exposed greater than risk in unexposed (positive association; possibly causal) If RR < 1 Risk in exposed less than risk in unexposed (negative association; possibly protective) Advantages of cohort studies Temporal relationship more certain Less opportunity for distortion of exposure data Can examine multiple disease outcomes Disadvantages of cohort studies Can be time consuming and expensive follow large group over long periods of time Potential bias due to drop outs from study Drop out in cohort study of oral health of older adults Baseline 3 years 7 years N=907 N=611 N=425 The Two Major Flavors of Cohort Studies: It s All in the Timing Concurrent Cohort Study Begun in 1995 Concurrent 1995 Defined Population Non-Randomized (Occurs naturally) 2005 Not 2015 6
Retrospective Cohort Study Begun in 1995 Defined Population Non-Randomized Retrospective 1975 Not 1985 1995 7